The background description is presented herein only by way of example to methods for employing semiconductor lasers to pump solid-state diode lasers. For example, in U.S. Pat. No. 3,982,201, Rosenkrantz et al. describe a solid-state laser that is pumped by single diodes or arrays of diodes to which the laser rod is directly end-coupled. Because the output wavelength of the diode laser array is a function of its temperature, the diode lasers are operated in a pulsed mode at a low duty cycle to maintain the array at a low enough temperature so that its output wavelength remains matched to the absorption bandwidth of the solid-state laser rod. The output power characteristics of this laser system are limited by the relatively inefficient match between the output of the diode lasers and mode volume of the solid-state laser rod.
In "Efficient LiNdP.sub.4 O.sub.12 Lasers Pumped with a Laser Diode," Applied Optics, vol. 18, No. 23 (Dec. 1, 1979), Kubodera and Otsuka describe the well-known practice of collecting the output light of a diode laser and focusing its expanded output light using conventional lenses, such as two microscope condenser lenses. This method is particularly well suited for applications where the emitter width and divergence of the diode laser are small. However, as the emitter dimensions and beam divergence increase, it becomes increasingly difficult to efficiently collect the output beam with collimating lens or lenses. It also becomes more difficult to focus the expanded beam into the solid-state laser crystal with sufficient depth of focus to allow efficient overlap of the pump beam throughout the resonator mode volume within the lasant.
In U.S. Pat. No. 4,710,940, Sipes, Jr. describes a neodymium:yttrium aluminum garnet (Nd:YAG) solid-state laser that is end-pumped by a diode laser array or by two diode laser arrays that have been combined by use of polarizing beam-splitting cubes. Sipes, Jr., cites the analysis of D. G. Hall in "Optimum Mode Size Criteria for Low Gain Lasers," Applied Optics, 1579-1583, vol. 20, (May 1, 1981), to suggest that the "pump profile shape does not matter much as long as all the pump light falls within the resonator mode." Sipes, Jr., notes, however, that Hall's analysis does not account for the divergence properties of Gaussian beams, so Sipes, Jr., suggests that, if required, the cross-section of the pump beam could be modified by use of a cylindrical lens.
In U.S. Pat. No. 4,761,786, Baer describes a Q-switched, solid-state laser that is end-pumped by a diode laser or diode laser array. The output light from the pump source is collected by a collimating lens and directed by a focusing lens to end-pump the laser rod. Baer notes that "other lenses to correct astigmatism may be placed between the collimating lens and focusing lens." Baer also describes an alternate embodiment that employs a remotely positioned diode laser pumping source coupled through an optical fiber, the output of which is focused via a lens into the laser rod.
In U.S. Pat. No. 4,763,975, Scifres et al. describe two optical systems that produce bright light output for a variety of applications, including pumping a solid-state laser such as a Nd:YAG. Scifres et al. describe an optical system that employs a plurality of diode lasers, each of which is coupled into one of a plurality of fiber-optic waveguides. The waveguides are arranged to form a bundle that delivers the light generated by the diode laser sources to the output end of the bundle. Optics, such as a lens, may be used to focus the light into a solid-state laser medium. Alternatively, the fiber bundle may be end-, and more specifically, "butt"-coupled to the laser rod (end-coupled and very close to or in contact with the laser rod).
Scifres et al. describe another optical system that employs a diode laser bar, broad-area laser, or other elongated source to pump a solid-state laser. The diode laser bar light output is coupled into a fiber-optic waveguide having an input end that has been squashed to be elongated and thereby have core dimensions and lateral and transverse numerical apertures that correspond respectively to those of emission dimension and lateral and transverse divergence angles of the laser bar. The output light from the fiber-optic waveguide is either focused using a lens into the end of the solid-state laser rod or butt-coupled to the rod. Scifres et al. state that either end of the fiber-optic waveguide can be curved. Although these methods attempt to match the output light from the fiber-optic waveguide to the resonant cavity mode of the solid-state laser, they are limited in efficiency by the numerical aperture of the sources that can be effectively collected and guided by the fiber-optic waveguides.
Solid-state lasers such as some of those described above may be employed in a variety of industrial operations including inspecting or microprocessing substrates such as electronic materials. The following description is presented herein only by way of example to Q-switched, diode-pumped, solid-state lasers employed for link processing during a semiconductor memory device repair operation.
For example, to repair a dynamic random access memory (DRAM), a first laser pulse is used to remove a conductive link to a faulty memory cell of a DRAM device, and then a second laser pulse is used to remove a resistive link to a redundant memory cell to replace the faulty memory cell. Because faulty memory cells needing link removals are randomly located, such laser repair processes are typically performed over a wide range of pulse repetition frequencies (PRFs), rather than at a constant PRF. This production technique is referred to in the industry as "on the fly" (OTF) link processing and allows for greater efficiency in the rate at which links on a given wafer can be repaired, thereby improving the efficiency of the entire DRAM production process.
The laser industry typically employs Q-switched, diode-pumped lasers (DPLs) using solid-state, neodymium-doped crystals to perform DRAM memory repair operations. DPLs are preferred over conventional arc-pumped Nd:YAG lasers for these operations because DPLs offer increased pumping source lifetimes. The typical mean time between failure (MTBF) is greater than 10,000 hours for diode-pumped lasers, whereas the typical MTBF is fewer than 1000 hours for arc-pumped lasers. In addition, DPLs do not require the water-cooling systems needed for arc-pumped lasers and are, therefore, better suited for operation in a clean-room environment. The previously described laser system of Baer is representative of such a laser system.
Laser energy per pulse typically decreases with increasing PRF, which is the inverse of the interpulse period (which is the length of time between the pulses emitted by the laser), and depend on the effective fluorescence lifetime of the metastable state of excited dopant or active ions in the crystal or lasant. For Q-switched, solid-state lasers, the laser pulse energies display a characteristic roll-off as the interpulse period decreases. This energy per pulse roll-off limits the upper PRF range for many laser memory repair processes. Each combination of epitaxial growth technique, material choice, and laser system design determines an acceptable range of deviation, often called a "process window" for pulse energies that will efficiently process the links. For many memory devices, the "process window" requires that laser pulse energy vary by less than 5% from a selected energy value.
Diode-pumped lasers, especially those employing neodymium:yttrium lithium fluoride (Nd:YLF) lasants, for example, are unable to limit pulse energy roll-off to less than 5% for PRFs&gt;1000 Hz because of the relatively long (500 .mu.sec) fluorescence lifetime of excited neodymium ions from the metastable state in a YLF lasant. DRAM memory repair applications typically employ a Nd:YLF lasant, however, because other useful lasant materials such as Nd:YAG are difficult to pump reliably using laser diodes and often exhibit pulse energy instability. The pulse energy instability can result from a mismatch between the spectral linewidths of the available high power, gain-guided aluminum gallium arsenide (AlGaAs) diode lasers and the absorption bandwidths of the most efficient lasants. The typical value of the full width, half-maximum spectral bandwidth for high power, gain-guided AlGaAs diode lasers is about 4 nm. Thus, while Nd:YAG has an absorption bandwidth of approximately 2 nm at 810 nm, Nd:YLF has a more closely matched absorption bandwidth of about 4 nm at 798 nm.
Neodymium:yttrium vanadate (Nd:YVO.sub.4) is another lasant that can be pumped by AlGaAs diode lasers. Nd:YVO.sub.4 has, however, a relatively low single ion fluorescence lifetime (typically less than 100 .mu.sec), thereby significantly limiting the energy per pulse that can be extracted from a Q-switched Nd:YVO.sub.4 laser. This energy per pulse is significantly less than that achievable from a Nd:YLF laser under similar pump power excitation levels. The limited energy per pulse of a Q-switched Nd:YVO.sub.4 laser can be a severe disadvantage whenever beam-shaping or other low-transmission laser optics are employed to direct the laser energy at, for example, metallic conductive links of a DRAM device.
Several neodymium-containing, stoichiometric, solid-state lasants such as lithium neodymium tetraphosphate (LNP) can also be pumped by AlGaAs diode lasers, as described by G. J. Dixon and L. S. Lingvay in "Close-Coupled Pumping of an Intracavity-Doubled Lithium Neodymium Tetraphosphate Laser," SPIE Solid-State Lasers, 291-293 vol. 1223 (1990). Stoichiometric neodymium compounds contain a very high neodymium concentration on the order of 10.sup.21 Nd/cm.sup.3 and thus have very high absorption coefficients. Solid-state laser crystals such as YLF, in which neodymium ions appear as dopant ions substituting for yttrium ions, have significantly lower neodymium concentrations, typically on the order of 10.sup.20 Nd/cm.sup.3.
Dixon and Lingvay employ the higher absorption coefficient of LNP in a diode-pumped LNP laser having a 0.4 mm-thick, LNP lasant crystal directly cemented to the submount of a Spectra Diode Laboratories Model 2240-H, phase-locked, diode laser array. This arrangement permits the LNP lasant to be thermoelectrically cooled to reduce the resonant loss and the energy transfer up-conversion from the laser metastable state that are typically exhibited by stoichiometric neodymium lasant compounds such as LNP. Nonstoichiometric neodymium laser compounds, such as Nd:YLF, do not typically exhibit these disadvantageous phenomena. It is also noted that the device described by Dixon and Lingvay does not include a Q-switch and does not attempt to collect the output of the diode-laser array via an optical system. The output of the diode laser is directly coupled into the LNP lasant crystal.
For the reasons set forth above, Nd:YLF is, therefore, an industry-preferred lasant for DRAM memory repair and silicon trimming operations, despite the limitations Nd:YLF places on the upper useful limit for PRF for OTF processing applications.
The power absorbed by a lasant is typically expressed as: EQU P.sub.z =P.sub.o e.sup.-.alpha.L, (1),
where P.sub.a is the power absorbed by the lasant; where P.sub.0 is the pumping power entering the lasant; where alpha is the absorption coefficient defined as .alpha.=4.pi.k/.lambda., where k is the extinction coefficient and .lambda. is the pumping wavelength; and where L is the length of the lasant. In conventional laser systems, L is much greater than or equal to 1/.alpha., the absorption depth of the lasant. Conventional nonstoichiometric lasants are fabricated to have length sufficient to absorb most of the energy coupled from a pumping source, even though the total percentage of excited active ions in the lasant mode volume decreases substantially as a function of the length of the lasant.
FIG. 1 shows the effect of PRF and effective fluorescence lifetime on energy per pulse of a Q-switched, solid-state laser. The data presented in FIG. 1 were determined in accordance with the theories of William G. Wagner and Bela A. Lengyel set out in "Evolution of the Giant Pulse in a Laser," Journal of Applied Physics, 2040-2046, vol. 34, No. 7, (July 1963) and of R. B. Chesler, M. A. Karr, and J. E. Geusic in "An Experimental and Theoretical Study of High Repetition Rate Q-Switched Nd:YAG Lasers," Proceedings of the IEEE, 1899-1914, vol. 58, No. 12 (Dec. 1970).
Mathematical equations that derive the dependence of the energy per pulse on the PRF assume that (1) the total number of excited ions is much smaller than the total number of active ions in the lasant and (2) the only depopulation mechanisms available to excited ions result from spontaneous and stimulated emission. These relationships are expressed as: EQU E.sub.1 =E.sub.0 (1-e.sup.-at),
where E.sub.0 represents the energy per pulse at very low PRFs (&lt;10 Hz); where E.sub.1 represents the energy per pulse at a PRF given by the inverse of the interpulse period t; and where the parameter ##EQU1## where .tau..sub.f is the single ion metastable state fluorescence lifetime of the active ion in the lasant and .omega..sub.p is the pumping rate, the effect of which is negligible at low pumping rates. The inverse of parameter a describes the effective fluorescence lifetime of the metastable state.